September marks not only the beginning of new academic year, but also the annual conference of the European Policy for Intellectual Property (EPIP) association. The 21st EPIP conference will take place in Turin, Italy, between 9 and 11 September 2026, and it will be hosted by Politecnico di Torino in cooperation with the Bordeaux School of Economics (Université de Bordeaux).
The conference brings together economics, law, management and political science scholars, stakeholders and policymakers to discuss challenges and policy issues linked to the evolution of the IP institutions in the context of the knowledge-based economy, this year under the theme of ‘Intellectual Property and Innovation Policy in the New Global Economy’.
By Fabio Fistarol via Unsplash
Our longstanding relationship with EPIP dates back over a decade. We were honoured to host the conference in Glasgow in 2015, when our Director Martin Kretschmer also served as the Association’s President. This year, Prof. Kretschmer will step down from the Board of EPIP, while CREATe Deputy Director Kristofer Erickson will step into the Board role.
A number of CREATe researchers are heading to Turin next week to discuss their ongoing research, keep an eye for us in the programme!
Ula Furgał & Amy Thomas: Redefining originality: between platforms and copyright
Content creators engage in creative labour. They take photos, write posts and share them with audiences on digital platforms. In principle such creative content can be copyrightable, especially considering the low originality threshold, and copyright’s ambivalence towards quality of protected work. Digital platforms became indispensable for both creation and distribution of creative content. While private actors, due to their powerful position can unilaterally determine the framework for creative production, encouraging and discouraging certain types of content and practices. One of the chief mechanisms of achieving this is through contract, including terms and conditions, terms of use, community guidelines and content policies.
Our work proposes to investigate the relationship between private ordering mechanisms of platforms and the key concept of originality in copyright. The research will employ discourse analysis methods to scrutinise the language of platforms’ terms to understand whether they replicate, redefine or substitute the concept of originality. In particular, we will analyse the use of alternative terms such as ‘quality’, ‘diversity’ or ‘authenticity’ used to describe content shared by creators. We will enquire as to what content is allowed, encouraged, or prohibited by the platforms, as well as what content is financially rewarded in programmes, such as TikTok’s ‘Creator Rewards Programme’. Indicatively, we propose to review terms for DeviantArt, Instagram, TikTok, Twitch and YouTube in a pilot study.
Our initial observation is that terms only recognise the negative aspects of copyright, with copyright infringement as a potential trigger for content moderation decisions, rather than to describe the quality of the creators work itself. Ultimately, this provokes the question as to whether platforms recognise and encourage creation of original content, or are more concerned with virality and using their power to shape creative practices and reinterpret basic copyright concepts.
Wednesday, 9 September 2026, P1-9i: Artificial Intelligence and Copyright, 5-7pm
Kristofer Erickson: Who’s in and who’s out? ‘NoAI’ opt-out behaviour on the DeviantArt image platform
DeviantArt is a social media platform and commercial marketplace for artists and illustrators. Artworks from the site have been used for AI model training without authorisation from individual users, including by ‘DreamUp’, the company’s own in-house generative AI system. Many users have expressed anger at the inclusion of their works in AI training sets, both internal and external. As a response, in 2022 DeviantArt rolled out a tagging system to enable creators to opt their works out of AI training using a setting available when uploading. In order to remain in compliance with DeviantArt’s updated terms of service (as well as the EU rights reservation under Art. 4(3) of the Copyright in the Digital Single Market Directive, (EU) 2019/790), third parties using DeviantArt-sourced content for AI training must ensure that their datasets exclude content that has the opt-out tags present.
This paper leverages an original dataset of 25,000 recently uploaded images (March-April 2026) along with their associated metadata to explore patterns in the prevalence of artists to opt out of AI training. We gather a diverse sample of images across the four main categories of ‘Digital Art’, ‘Fantasy’, ‘Traditional Art’, and ‘3D Art’. For each image, we record variables related to the artist: account age, account status (basic or professional), gender, territory, and quantity of previous uploads. We further observe the platform performance (downloads, favourites) to assess whether there is a relationship between opt-out and eventual popularity or performance of individual artworks. The findings have relevance for copyright policy by providing a view into the prevalence and patterns of opt-out behaviour by artists. Our study has further relevance for questions for content diversity and the wider effects of limitations on AI training, by assessing whether opt out may privilege or bias datasets towards certain groups of works (subject matter, quality, commercial potential, etc.).
Friday, 11 September 2026, P5-5i: AI Creativity, Content Licensing, and Platform Authorship, 2-4pm
Martin Kretschmer & Amy Thomas: Towards an AI Licensing economy? Empirical evidence on agreements between creative content firms and AI developers
Following the public release of ChatGPT in November 2022, policy and litigation pressure on AI developers increased dramatically, in particular with respect to the data used for model training. This paper reports empirical evidence from all publicly reported commercial agreements between creative content firms and AI developers, covering the period from October 2022 to January 2026. The underlying data (n = 274 deals) has been made available open access by the CREATe centre (https://www.create.ac.uk/blog/2025/02/24/the-ai-licensing-economy/).
Agreements were coded for parties (parent companies and subsidiaries), modality (text, image, audio, video), sectors (including press, scientific content, user generated content, visual art etc), model compute, deal type (including partnership, content licence), deal size, and legal clauses (such as exclusivity, time, revenue share).
Preliminary findings include:
AI licensing is increasing and accelerating but Big Tech is not the biggest licensee. Perplexity, Prorata and OpenAI make for the largest share of known commercial agreements, whereas older ‘Big Tech’ companies, such as Amazon, Apple, Google, Meta, and Microsoft, have comparatively fewer agreements. Reasons may include that these companies already have privileged access to data.
The News/Media sector account for the largest share of licences agreed. 62% of known commercial agreements of AI developers are with News or Media companies. This is in line with expectations that up-to-date information is needed to offer accurate responses to user queries.
Public reports are often insufficient to characterise the legal character of the agreements. They can range from data access agreements, to copyright licences, from assurances not to sue, to comprehensive partnerships and equity stakes. Agreements may involve training and output activities, and cover various technologies, such as augmented retrieval. They may be the result of out-of-court settlement of litigation.
Overall, the data indicates that the AI licensing economy is already happening at scale. Reasons for agreements seem to include copyright risk, technological constraints (e.g. API access for augmented retrieval), and the need for quality data for specific tailored new services.
Friday, 11 September 2026, P5-5i: AI Creativity, Content Licensing, and Platform Authorship, 2-4pm
Phalguni Mahapatra: Beyond Repair: How Private Ordering Can Limit Innovation and User Creativity
This research examines informal repair practices within the mobile device ecosystem in India, focusing on how repair operates in contexts shaped by economic constraints, uneven access to technology, and increasingly restrictive product design architectures. Situating repair within broader debates on the Right to Repair, the study highlights how its meaning and practice differ across socio-economic contexts, particularly between the Global North and the Global South. It further engages with emerging discussions on intellectual property (IP) and open innovation, exploring how contemporary IP regimes and technological protection measures shape, and often constrain, access to repair-related knowledge, tools, and components.
The study investigates the organisation of grassroots repair networks, the role of reuse and refurbished economies, and the circulation of knowledge within informal ecosystems. It conceptualises repair not merely as maintenance but as a potential site of user-led and practice-based innovation, while remaining attentive to how such practices operate within conditions of legal ambiguity and technological restriction. A comparative dimension between India and Glasgow is employed to examine how differing regulatory, economic, and cultural environments influence repair practices and access to technological systems.
Methodologically, the research adopts a qualitative socio-legal approach based on ongoing fieldwork. This includes semi-structured interviews with independent repairers, ethnographic observations of repair practices, and engagement with informal repair markets and supply chains. By focusing on everyday repair activities and localised knowledge systems, the study seeks to understand how technological engagement is shaped in practice and how repair functions within broader socio-economic and regulatory environments. The findings aim to contribute to debates on IP, innovation governance, and technological access by foregrounding the role of informal actors in sustaining and reconfiguring contemporary technology ecosystems.
Friday, 11 September 2026, P5-7i: Generative AI, Human Agency, and Private Ordering, 2-4pm